How to integrate a Data Science Team to my company
As our Data Science Team and our company needs grow, it’s necessary to create a completely new department that must be organized, controlled, monitored and administrated. This big organizational change suggests that a new group must have more established roles and responsibilities. This must all be in relation to other projects and facilities. Then, how do I integrate a Data Science Team to my company? in other words, how do you integrate data scientists to your company?
The following are the most used models:
– Consulting model
Work is done on teams, but their function inside an organization is consultancy, so that different company areas can claim them for specific tasks. The consulting model is the most adequate model for small companies with sporadic data science tasks and from small to medium scale.
Some drawbacks of the consulting model:
- Firstly, bad data quality can happen to be a fundamental defect for this model. As the team cannot stick to the best practices for each task, they have to sacrifice quality for commercial needs that require quick solutions.
- Also, we can find low levels of motivation. As the area is not completely involved in product creation and decision making, they might have low interest in the result.
- A serious problem of a consulting model is uncertainty. Time periods are not clear when the team is clearly not familiarized with data sources and their context of appearance. Complex and long term projects are not very accessible, as sometimes, specialists work many years over the same problems in order to achieve great results.
- The priority method is also not clear. It’s difficult to identify when a data science administrator prioritizes and assigns tasks for people involved and what objectives to put in first place.
– Centralized model
This structure allows using analytics in strategical tasks: one team serves to the entire organization in various projects. Not only does it provide the team with a long term vision and better management of their resources, but it also encourages professional growth. The only problem is the danger of transforming an analytical function into a support one.
One of the best use cases in order to create a specialized team is when the analysis demand and the number of analysts grows rapidly. This requires an urgent assignation of this resources. By introducing a centralized approach, a company indicates that they consider data as a strategical concept and that it’s ready to build a department or area of analysis that is equivalent to sales or marketing.
Some drawbacks of the centralized model:
- There’s a high probability of getting isolated and facing disconnection between a data analysis team and business units. As the data analysis team does not participate in usual tasks of real value business units, it’s possible that they are not familiar with the needs and problems of them. This might lead to a limited recommendation relevance. Because of this, they may not be used or get ignored.This conducts to challenges in significate cooperation with a product team. Once the analysis group has found a way to deal with a problem, they suggest a solution to the product team. The biggest problem is that this solution might not fit into the product’s roadmap and problems may arise. The only solution is creating a team to evaluate, design and implement the suggested solution. However, this alternative requires assigning plenty of resources.
– Center of excellence model .CoE
The centralized focus is maintained with only one coordination center, but teams are assigned to different units of the organization. This is the most balanced structure, where the analysis activities are highly coordinated, but business unit’s experts won’t be removed.
As the interactions are well balanced, this model is being chosen by more and more people, especially for company like organizations. It works much better for companies with a corporate strategy and a completely developed roadmap.
Some of the drawbacks of the excellence model are:
- Even though this model is balanced, there isn’t only one centralized group that deals with company-level problems. Every group would be solving problems in their units
- Another drawback is that there’s not a unit of innovation. A group of specialists that focus mainly in the avant-garde solutions and long term data initiatives instead of day to day needs.
– Federal Model
This model is relevant when there’s a bigger requirement of analytical talent in the whole company. It involves a specialized team or analysis group that works from a main point and boards complex and multifunctional tasks. The rest of the teams are distributed in the same way as the Center of Excellence Model.
The Federal Model works better for companies where processes and analysis tasks have a systemic nature and need daily updates. This model can be useful for company level objectives as well as dashboard design and tailored analysis with different modelling types.
Some drawbacks of the Federal Model:
- Acquisition expenses and talent retention: This model suggests an independent specialist for each team and central data management. This is why this model is not the best option for companies that are on their first steps of analysis adoption.
- Crossed functionality can create a problematic context. It can be missing a power parity among all leadership roles in the team and cause late hand ins or questionable results because of the constant conflicts between the unit team leaders and the CoE administration’s leaders
– Democratic Model
This model is an additional way to think about data culture. It implies that everyone in the organization has access to data via BI tools or other tools. This means that it can be combined with any other model, it can have a Federated focus with COE and analysis specialists in each area and at the same time, expose BI tools to anyone interested in using data for their functions, which is excellent in terms of encouraging data culture.
Drawbacks of the democratic model:
- A lot is invested in infrastructure, tools and training in data science
- More people will be needed in order to avoid responsibilities overlapping, allowing people involved to dedicate to their speciality and defined role in the team
Which model can be better adapted to your company?
Cómo integrar un Data Science Team a mi compañía